mlsanityData to 14 Oct 2026

Mario Plays on a Manifold: Generating Functional Content in Latent Space through Differential Geometry

arXiv 2206.00106 ↗First posted 31 May 2022cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

Deep generative models can automatically create content of diverse types. However, there are no guarantees that such content will satisfy the criteria necessary to present it to end-users and be functional, e.g. the generated levels could be unsolvable or incoherent. In this paper we study this problem from a geometric perspective, and provide a method for reliable interpolation and random walks in the latent spaces of Categorical VAEs based on Riemannian geometry. We test our method with "Super Mario Bros" and "The Legend of Zelda" levels, and against simpler baselines inspired by current practice. Results show that the geometry we propose is better able to interpolate and sample, reliably staying closer to parts of the latent space that decode to playable content.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 29 of its 39 references that are in the map. First posted on arXiv 31 May 2022.

Builds on

29 of 39 references.

References · 29

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  3. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 6,013
  4. 2013Auto-Encoding Variational Bayescited by 4,053
  5. 2017Latent Space Oddity: on the Curvature of Deep Generative Modelscited by 174
  6. 2017Procedural Content Generation via Machine Learning (PCGML)cited by 137
  7. 2018Evolving Mario Levels in the Latent Space of a Deep Convolutional Generative Adversarial Networkcited by 76
  8. 2020Mat'ern Gaussian Processes on Graphscited by 66
  9. 2017Metrics for Deep Generative Modelscited by 53
  10. 2016The VGLC: The Video Game Level Corpuscited by 41
  11. 2016Super Mario as a String: Platformer Level Generation Via LSTMscited by 40
  12. 2019Bootstrapping Conditional GANs for Video Game Level Generationcited by 35
17 more
  1. 2021The Geometry of Deep Generative Image Models and its Applicationscited by 35
  2. 2021Illuminating Diverse Neural Cellular Automata for Level Generationcited by 35
  3. 2014Metrics for Probabilistic Geometriescited by 28
  4. 2021Generating and Blending Game Levels via Quality-Diversity in the Latent Space of a Variational Autoencodercited by 27
  5. 2019Fast and Robust Shortest Paths on Manifolds Learned from Datacited by 25
  6. 2019Generative Models for Automatic Chemical Designcited by 23
  7. 2021Learning Riemannian Manifolds for Geodesic Motion Skillscited by 23
  8. 2021Pulling back information geometrycited by 23
  9. 2020Interactive Evolution and Exploration Within Latent Level-Design Space of Generative Adversarial Networkscited by 22
  10. 2020TOAD-GAN: Coherent Style Level Generation from a Single Examplecited by 22
  11. 2020CPPN2GAN: Combining Compositional Pattern Producing Networks and GANs for Large-scale Pattern Generationcited by 19
  12. 2020Video Game Level Repair via Mixed Integer Linear Programmingcited by 18
  13. 2020Conditional Level Generation and Game Blendingcited by 17
  14. 2018DOOM Level Generation using Generative Adversarial Networkscited by 15
  15. 2020Sequential Segment-based Level Generation and Blending using Variational Autoencoderscited by 14
  16. 2021Dungeon and Platformer Level Blending and Generation using Conditional VAEscited by 10
  17. 2020Game Level Clustering and Generation using Gaussian Mixture VAEscited by 9

Built on by

Cited by year

Citing papers per year: 2022: 1; 2023: 2; 2024: 0; 2025: 2.

5 papers.

  1. 2023MarioGPT: Open-Ended Text2Level Generation through Large Language Modelscited by 64
  2. 2022Identifying latent distances with Finslerian geometrycited by 3
  3. 2023Tools for Landscape Analysis of Optimisation Problems in Procedural Content Generation for Gamescited by 2
  4. 2025Analysis of Robustness of a Large Game Corpuscited by 1
  5. 2025PepCompass: Navigating peptide embedding spaces using Riemannian Geometrycited by 0